|
| 1 | +--- |
| 2 | +title: "admiral 1.0.0" |
| 3 | +author: |
| 4 | + - name: Ben Straub |
| 5 | +description: "1.0.0 brings a commitment to stability, new features, a few bug fixes, argument alignment and onboarding resources!" |
| 6 | +date: "2024-01-04" |
| 7 | +# please do not use any non-default categories. |
| 8 | +# You can find the default categories in the repository README.md |
| 9 | +categories: [admiral, ADaMs] |
| 10 | +# feel free to change the image |
| 11 | +image: "new_features.jpg" |
| 12 | + |
| 13 | +--- |
| 14 | + |
| 15 | +<!--------------- typical setup -----------------> |
| 16 | + |
| 17 | +```{r setup, include=FALSE} |
| 18 | +long_slug <- "2023-12-18_admiral_1_0" |
| 19 | +# renv::use(lockfile = "renv.lock") |
| 20 | +``` |
| 21 | + |
| 22 | +<!--------------- post begins here -----------------> |
| 23 | + |
| 24 | +[`admiral 1.0.0`](https://pharmaverse.github.io/admiral/index.html) is out on [CRAN](https://cran.r-project.org/web/packages/admiral/index.html). This release brings several new features to your tool set |
| 25 | +for working with ADaMs in R. [`admiral 1.0.0`](https://pharmaverse.github.io/admiral/index.html) also brings needed stability to users who were wishing |
| 26 | +to adopt [admiral](https://pharmaverse.github.io/admiral/index.html), but were a little worried by the fast deprecation and experimentation for `pre-v1.0.0` releases. |
| 27 | + |
| 28 | +This blog post will discuss our commitment to stability, walk you through the new features available, discuss some of the bug fixes, a push for common APIs across our functions, and showcase the resources available to help you on-board to [admiral](https://pharmaverse.github.io/admiral/index.html). |
| 29 | + |
| 30 | +# Commitment to Stability |
| 31 | + |
| 32 | +`{admiral}` was born out of a conversation between Thomas Neitmann and Michael Rimler almost three years ago. Check out this [history blog post](https://pharmaverse.github.io/blog/posts/2023-10-10_pharmaverse_story/pharmaverse_story.html) for some of additional context on its development. Over the last three years, we have worked at a furious pace to produce functions that can do almost all derivations needed for common ADaMs. However, we have noticed that some functions are very similar to others and with some minor tweaks could be made redundant in favor of other functions. This has lead to some deprecation of either whole functions or arguments as we have sought to consolidate. Unfortunately, this creates a lot of frustration for users. |
| 33 | + |
| 34 | +Enter Superseded!! |
| 35 | + |
| 36 | +Now with the release of `1.0.0` we are going to _supersede_ functions going forward. This is the process where when we introduce new functions to replace other functions, we will no longer fast-deprecate the replaced functions. Now we will supersede them, i.e. we will keep the function in the codebase for a long time, but recommend the use of the new function. If the function were to be removed from the codebase this will be done many years from the time it is superseded. Please note that although superseded functions are not in scope for active development, any bugs identified within them will nevertheless be fixed. |
| 37 | + |
| 38 | + |
| 39 | +# New Features |
| 40 | + |
| 41 | +We have 3 new functions available in this release: |
| 42 | + |
| 43 | +* [derive_vars_extreme_event()](https://pharmaverse.github.io/admiral/reference/derive_vars_extreme_event.html) |
| 44 | +* [derive_var_merged_ef_msrc()](https://pharmaverse.github.io/admiral/reference/derive_var_merged_ef_msrc.html) |
| 45 | +* [derive_vars_computed()](https://pharmaverse.github.io/admiral/reference/derive_vars_computed.html) |
| 46 | + |
| 47 | +and a new option in [derive_var_atoxgr_dir()](https://pharmaverse.github.io/admiral/reference/derive_var_atoxgr_dir.html) and [derive_var_anrind()](https://pharmaverse.github.io/admiral/reference/derive_var_anrind.html) to handle floating point issues. |
| 48 | + |
| 49 | +## `derive_vars_extreme_event()` |
| 50 | + |
| 51 | +This function takes available records from user-defined events by selecting the extreme observations and appending them as a variable(s) to your dataset. [derive_vars_extreme_event()](https://pharmaverse.github.io/admiral/reference/derive_vars_extreme_event.html) works similar to [`derive_extreme_event()`](https://pharmaverse.github.io/admiral/reference/derive_extreme_event.html), but instead of adding observations the function will add variable(s). |
| 52 | + |
| 53 | +Let's take a peek with a very simple example where we just use ADSL! The documentation for |
| 54 | +[derive_vars_extreme_event()](https://pharmaverse.github.io/admiral/reference/derive_vars_extreme_event.html) has a much richer example with events from other domains that is more aligned to where you would use this function. |
| 55 | + |
| 56 | +Let us make some dummy `ADSL` data and load up our packages. The goal here is to add two new variables `LSTALVDT` and `DTHFL` based on a list of objects that are used to specify the following: |
| 57 | + |
| 58 | + - the dataset to look at |
| 59 | + - a set of conditions |
| 60 | + - what to set the values for the new variables |
| 61 | + |
| 62 | +```{r, message = FALSE, warning = FALSE} |
| 63 | +library(tibble) |
| 64 | +library(admiral) |
| 65 | +library(lubridate) |
| 66 | +
|
| 67 | +adsl <- tribble( |
| 68 | + ~STUDYID, ~USUBJID, ~TRTEDT, ~DTHDT, |
| 69 | + "PILOT01", "01-1130", ymd("2014-08-16"), ymd("2014-09-13"), |
| 70 | + "PILOT01", "01-1133", ymd("2013-04-28"), ymd(""), |
| 71 | + "PILOT01", "01-1211", ymd("2013-01-12"), ymd(""), |
| 72 | + "PILOT01", "09-1081", ymd("2014-04-27"), ymd(""), |
| 73 | + "PILOT01", "09-1088", ymd("2014-10-09"), ymd("2014-11-01"), |
| 74 | +) |
| 75 | +``` |
| 76 | + |
| 77 | + |
| 78 | + |
| 79 | +In this example, we only use `ADSL` as the source dataset, so it is a bit contrived, but much more compact for us. Note the use of the `events` that is taking in our list of event objects and the different conditions and values we set to create our `LSTALVDT` and `DTHFL` variables. |
| 80 | + |
| 81 | +```{r} |
| 82 | +derive_vars_extreme_event( |
| 83 | + adsl, |
| 84 | + by_vars = exprs(STUDYID, USUBJID), |
| 85 | + events = list( |
| 86 | + event( |
| 87 | + dataset_name = "adsl", |
| 88 | + condition = !is.na(DTHDT), |
| 89 | + set_values_to = exprs(LSTALVDT = DTHDT, DTHFL = "Y") |
| 90 | + ), |
| 91 | + event( |
| 92 | + dataset_name = "adsl", |
| 93 | + condition = !is.na(TRTEDT), |
| 94 | + set_values_to = exprs(LSTALVDT = TRTEDT, DTHFL = "N") |
| 95 | + ) |
| 96 | + ), |
| 97 | + source_datasets = list(adsl = adsl), |
| 98 | + order = exprs(LSTALVDT), |
| 99 | + mode = "last", |
| 100 | + new_vars = exprs(LSTALVDT = LSTALVDT, DTHFL = DTHFL) |
| 101 | +) |
| 102 | +``` |
| 103 | + |
| 104 | +Okay! We used a very small example to showcase how to find extreme observations and |
| 105 | +appending this information as new variables to our `ADSL` dataset. Highly recommend checking |
| 106 | +out the more detailed example in the [function documentation](https://pharmaverse.github.io/admiral/reference/derive_vars_extreme_event.html) to see its true power! |
| 107 | + |
| 108 | +## `derive_var_merged_ef_msrc()` |
| 109 | + |
| 110 | +This function has some similarity to [derive_vars_extreme_event()](https://pharmaverse.github.io/admiral/reference/derive_vars_extreme_event.html), but now we are only looking at adding a single flag variable based on checking conditions across multiple datasets. |
| 111 | + |
| 112 | +We develop some _simple_ dummy data for `ADSL`, `CM` and `PR`. Our goal is to flag patients who have `CMCAT = "ANTI-CANCER"` in the `CM` dataset or have records in the `PR` dataset. Any participants who meet these conditions will have our new variable `CANCTRFL` set as `"Y"`. |
| 113 | + |
| 114 | +```{r, message = FALSE, warning = FALSE} |
| 115 | +adsl <- tribble( |
| 116 | + ~USUBJID, |
| 117 | + "1", |
| 118 | + "2", |
| 119 | + "3", |
| 120 | + "4", |
| 121 | +) |
| 122 | +
|
| 123 | +cm <- tribble( |
| 124 | + ~USUBJID, ~CMCAT, ~CMSEQ, |
| 125 | + "1", "ANTI-CANCER", 1, |
| 126 | + "1", "GENERAL", 2, |
| 127 | + "2", "GENERAL", 1, |
| 128 | + "3", "ANTI-CANCER", 1 |
| 129 | +) |
| 130 | +
|
| 131 | +pr <- tribble( |
| 132 | + ~USUBJID, ~PRSEQ, |
| 133 | + "2", 1, |
| 134 | + "3", 1 |
| 135 | +) |
| 136 | +``` |
| 137 | + |
| 138 | + |
| 139 | +Now we have the argument `flag_events` that takes a list of objects where we define the conditions and datasets to check in. |
| 140 | + |
| 141 | +```{r} |
| 142 | +derive_var_merged_ef_msrc( |
| 143 | + adsl, |
| 144 | + flag_events = list( |
| 145 | + flag_event( |
| 146 | + dataset_name = "cm", |
| 147 | + condition = CMCAT == "ANTI-CANCER" |
| 148 | + ), |
| 149 | + flag_event( |
| 150 | + dataset_name = "pr" |
| 151 | + ) |
| 152 | + ), |
| 153 | + source_datasets = list(cm = cm, pr = pr), |
| 154 | + by_vars = exprs(USUBJID), |
| 155 | + new_var = CANCTRFL |
| 156 | +) |
| 157 | +``` |
| 158 | + |
| 159 | +Let's go! We searched over multiple datasets, `CM` and `PR`, with multiple conditions and appended a new variable `CANCTRFL` to `ADSL` setting to `"Y"` if those conditions were met. |
| 160 | + |
| 161 | +## `derive_vars_computed()` |
| 162 | + |
| 163 | +This function is very similar to [derive_param_computed()](https://pharmaverse.github.io/admiral/reference/derive_param_computed.html), but instead of adding observations we are going to add variable(s). Very handy when wanting to add some additional variables to `ADSL`, e.g. baseline variables. |
| 164 | + |
| 165 | +Let's make some dummy data for an `ADSL` and `ADVS`. Our goal is to derive a `BMIBL` variable pulled from `ADVS` and append to `ADSL`. |
| 166 | + |
| 167 | +```{r} |
| 168 | +adsl <- tribble( |
| 169 | + ~STUDYID, ~USUBJID, ~AGE, ~AGEU, |
| 170 | + "PILOT01", "01-1302", 61, "YEARS", |
| 171 | + "PILOT01", "17-1344", 64, "YEARS" |
| 172 | +) |
| 173 | +
|
| 174 | +advs <- tribble( |
| 175 | + ~STUDYID, ~USUBJID, ~PARAMCD, ~PARAM, ~VISIT, ~AVAL, ~AVALU, ~ABLFL, |
| 176 | + "PILOT01", "01-1302", "HEIGHT", "Height (cm)", "SCREENING", 177.8, "cm", "Y", |
| 177 | + "PILOT01", "01-1302", "WEIGHT", "Weight (kg)", "SCREENING", 81.19, "kg", "N", |
| 178 | + "PILOT01", "01-1302", "WEIGHT", "Weight (kg)", "BASELINE", 82.1, "kg", "Y", |
| 179 | + "PILOT01", "01-1302", "WEIGHT", "Weight (kg)", "WEEK 2", 81.19, "kg", "N", |
| 180 | + "PILOT01", "01-1302", "WEIGHT", "Weight (kg)", "WEEK 4", 82.56, "kg", "N", |
| 181 | + "PILOT01", "01-1302", "WEIGHT", "Weight (kg)", "WEEK 6", 80.74, "kg", "N", |
| 182 | + "PILOT01", "17-1344", "HEIGHT", "Height (cm)", "SCREENING", 163.5, "cm", "Y", |
| 183 | + "PILOT01", "17-1344", "WEIGHT", "Weight (kg)", "SCREENING", 58.06, "kg", "N", |
| 184 | + "PILOT01", "17-1344", "WEIGHT", "Weight (kg)", "BASELINE", 58.06, "kg", "Y", |
| 185 | + "PILOT01", "17-1344", "WEIGHT", "Weight (kg)", "WEEK 2", 58.97, "kg", "N", |
| 186 | + "PILOT01", "17-1344", "WEIGHT", "Weight (kg)", "WEEK 4", 57.97, "kg", "N", |
| 187 | + "PILOT01", "17-1344", "WEIGHT", "Weight (kg)", "WEEK 6", 58.97, "kg", "N" |
| 188 | +) |
| 189 | +``` |
| 190 | + |
| 191 | +Take a look at how we use `new_vars` and `filter_add`. We use a function inside of `new_vars` to help us calculate the `BMI` while using the `filter_add` argument to only look at baseline records for the calculation. |
| 192 | + |
| 193 | +```{r} |
| 194 | +derive_vars_computed( |
| 195 | + dataset = adsl, |
| 196 | + dataset_add = advs, |
| 197 | + by_vars = exprs(STUDYID, USUBJID), |
| 198 | + parameters = c("WEIGHT"), |
| 199 | + constant_by_vars = exprs(STUDYID, USUBJID), |
| 200 | + constant_parameters = c("HEIGHT"), |
| 201 | + new_vars = exprs(BMIBL = compute_bmi(height = AVAL.HEIGHT, weight = AVAL.WEIGHT)), |
| 202 | + filter_add = ABLFL == "Y" |
| 203 | +) |
| 204 | +``` |
| 205 | + |
| 206 | +Alright! Simple enough. We just took records from `ADVS`to help us calculate the `BMI` at baseline using this function and appended our new variable to `ADSL`. |
| 207 | + |
| 208 | +# Argument Alignment and Bug Fixes |
| 209 | + |
| 210 | +## Argument Alignment |
| 211 | + |
| 212 | +A huge push was made for `1.0.0` to help align our arguments across all of `{admiral}` functions. What does this mean? We identified arguments in functions where the argument did the same things but was slightly named differently. For `1.0.0`, we really want users to have a solid API for `{admiral}` functions. |
| 213 | + |
| 214 | +Let's take a peak at the function [`consolidate_metadata()`](https://pharmaverse.github.io/admiral/reference/consolidate_metadata.html) to even better understand our goal here. |
| 215 | + |
| 216 | +``` |
| 217 | +consolidate_metadata( |
| 218 | + datasets, |
| 219 | + key_vars, |
| 220 | + source_var = SOURCE, |
| 221 | + check_vars = "warning", |
| 222 | + check_keys, |
| 223 | + check_type = "error" |
| 224 | +) |
| 225 | +``` |
| 226 | + |
| 227 | +In previous versions of `{admiral}` the `consolidate_metadata()` function had the argument `check_keys`, which helps to check uniqueness. Other functions had a similar argument, but were called `check_unique`. Therefore, to better align our common API for `{admiral}` functions we decided to rename the `check_keys` argument to `check_unique`. You can follow the discussion around this renaming effort in this [GitHub Issue](https://github.com/pharmaverse/admiral/issues/2184). |
| 228 | + |
| 229 | +```{r, fig.align = 'center', echo=FALSE} |
| 230 | +knitr::include_graphics("check.png") |
| 231 | +``` |
| 232 | + |
| 233 | + |
| 234 | +The argument has a deprecated tag in the function documentation and will issue a warning to users. There was quite a bit of renaming of arguments for `1.0.0` so there are quite a few of these tags in our documentation. In subsequent releases, these arguments will be removed. Please see the [changelog](https://pharmaverse.github.io/admiral/news/index.html#admiral-100) if you would like to explore other functions that had arguments renamed. The issues are linked to each rename so you can follow along with the discussions! |
| 235 | + |
| 236 | +## Bug Fixes |
| 237 | + |
| 238 | +We love fixing bugs and take them incredibly seriously - especially when identified by members from the community. |
| 239 | + |
| 240 | +If you find a pesky bug, please fill out a [Bug Report](https://github.com/pharmaverse/admiral/issues/new/choose) on our Issues Tab. |
| 241 | + |
| 242 | +Each bug fixed by our development team is documented in our [changelog](https://pharmaverse.github.io/admiral/news/index.html#updates-of-existing-functions-1-0-0) with the Issue linked. |
| 243 | + |
| 244 | +```{r, fig.align = 'center', echo=FALSE} |
| 245 | +knitr::include_graphics("issue_tag.png") |
| 246 | +``` |
| 247 | + |
| 248 | + |
| 249 | +For example, if you click through the issue for [`derive_extreme_event()`](https://github.com/pharmaverse/admiral/issues/2291) that identified a problem where the `condition` was ignored if the `mode` argument was used, you can see the Bug Report along with a reproducible example. You can also see the Pull Request for the exact code changes that are addressing this bug linked in the Issue! Way cool! |
| 250 | + |
| 251 | + |
| 252 | +# New On-boarding Resources |
| 253 | + |
| 254 | +`{admiral}` has a lot of functions for working with ADaMs. This can be overwhelming for new users and we really sympathize. To help new users onboarding to `{admiral}` we have developed the following resources: |
| 255 | + |
| 256 | +## [admiraldiscovery](https://pharmaverse.github.io/admiraldiscovery/index.html) |
| 257 | + |
| 258 | +This is a dedicated website that lists out in a tabular format standard ADaM datasets and their common variables with corresponding `{admiral}` functions that could be used to create the variables. Very handy when you just want to get some starter code on deriving `EOSDT` or `TRTSDT`! |
| 259 | + |
| 260 | + |
| 261 | +```{r, fig.align = 'center', echo=FALSE} |
| 262 | +knitr::include_graphics("admiraldisc.png") |
| 263 | +``` |
| 264 | + |
| 265 | + |
| 266 | +## [admiral Cheat Sheet](https://github.com/pharmaverse/admiral/blob/main/inst/cheatsheet/admiral_cheatsheet.pdf) |
| 267 | + |
| 268 | +Inspired by other R package cheat sheets! We try and surface commonly needed functions for doing ADaM derivations with simple tables to show how the data is transforming. |
| 269 | + |
| 270 | +```{r, fig.align = 'center', echo=FALSE} |
| 271 | +knitr::include_graphics("cheatsheet.png") |
| 272 | +``` |
| 273 | + |
| 274 | +## Way Back Machine |
| 275 | + |
| 276 | +Studies can last a long time. Adopting R as your primary analysis language for your study can introduce certain risks around package dependencies. Fixing those dependencies to certain package versions can help mitigate those risks. Unfortunately, as package websites are updated those helpful documents, examples and vignettes can be lost as the version changes. Do not lose heart `{admiral}` users. If you decided to fix to a certain version of `{admiral}`, we have you covered with our __Way Back Machine__ that allows you to change the website documentation back to the version you are using. |
| 277 | + |
| 278 | +```{r, fig.align = 'center', echo=FALSE} |
| 279 | +knitr::include_graphics("wayback.png") |
| 280 | +``` |
| 281 | + |
| 282 | +<!--------------- appendices go here -----------------> |
| 283 | + |
| 284 | +```{r, echo=FALSE} |
| 285 | +source("appendix.R") |
| 286 | +insert_appendix( |
| 287 | + repo_spec = "pharmaverse/blog", |
| 288 | + name = long_slug |
| 289 | +) |
| 290 | +``` |
0 commit comments